NAVI-Orbital: First In-Orbit Demonstration of a Zero-Shot Vision-Language Model for Autonomous Earth Observation Enhancing Pathological VLMs with Cross-scale Reasoning Where Did the Variability Go? From Vibe Coding to Product Lines by Regeneration A Hybrid LSTM–Vision Transformer Architecture for Predicting HRRR Forecast Errors Calibrated Sampling-Free Uncertainty Estimation in Bayesian Deep Learning